Dynamically building an unstructured information management architecture (UIMA) pipeline

摘要

A pipeline development environment includes a toolset that includes a visual design editor. The editor comprises a display interface having a palette of known Annotators that may be selected by a developer. The pipeline development environment also includes or has associated therewith a data repository. The data repository stores datasets. A particular dataset is associated with an Annotator and comprises dependency data generated from execution of a pipeline (or some portion thereof). The repository typically stores datasets from many pipeline runs, including runs of other pipelines, multiple runs of a given pipeline with different inputs, etc. Using the editor, a developer creates a visual representation of the pipeline. As Annotators are added into the pipeline, system tooling dynamically generates the descriptor files and other configuration parameters (for the new pipeline), preferably based on the dependency data associated with the individual Annotators and retrieved from the repository.